Principal Component Analysis in various Languages

pca.cpp

#include <iostream>
#include <dlib/statistics/dpca.h>
#include <dlib/statistics/dpca_abstract.h>
#include <dlib/matrix.h>
#include <initializer_list>


using namespace dlib;

int main()
{

    matrix< double , 2,1 > vectors_for_pca[]
        {
            { 4.0 , 3.0  } ,
            { 3.0 , 2.25  } ,
            { 2.0 , 1.5  } ,
            { 1.03 , 0.7533  } ,
        };


    //std::cout << "data:\nNR=" << vectors_for_pca.nr() << "\nNC=" << vectors_for_pca.nc() << "\n";
    dlib::discriminant_pca< matrix< double>  > pca;
    for (int i=0 ; i< 4 ;++i) {
        pca.add_to_total_variance( vectors_for_pca[i] );
    }

    matrix<double> pca_out(2,2);
    matrix<double> pca_evout(2,1);

    // the third value acts as a threshold for filtering out the eigenvalues
    // 1.0 --> return all values
    // 0.9 --> return values until the cumulative sum of eigenvalues (when sorted descending) > 90% total eigen values
    pca.dpca_matrix(pca_out, pca_evout,0.99);

    // in vector size is the number if fields in each entry
    auto vs = pca.in_vector_size();
    std::cout << "in_vector_size: " << vs <<
                 "\nprincipal_transformation_matrix: \n" << pca_out <<
                 "\neigenvalues:\n " << pca_evout << "\n";

    std::cout << "eigen value matrix of size 2:\n";
    auto m = pca.dpca_matrix_of_size(2);
    std::cout << m;


    return 0;
}

pca.py

import numpy as np
from sklearn.decomposition import PCA

p = PCA( 2 )

arr = [
   [ 4 , 3 ],
   [ 3 , 2.25 ] ,
   [ 2 , 1.5 ] ,
   [ 1 , 0.75 ] ]

p.fit( arr )

# eigenvalues..
print p.explained_variances_

# eigenvectors..
print p.components_